On 8/26/24 10:45 AM, Segher Boessenkool wrote:
> On Thu, Aug 22, 2024 at 08:48:19PM -0500, Peter Bergner wrote:
>> I agree, there probably is code in the backend that currently handles TImode
>> that should probably be changed to handle both by using your new macro.
>
> That is what mode classes are for, or just the mode size here :-)
I don't think mode size alone work work here, since that would include
TF/KF/IF modes. Mode class & mode size would probably work, but I don't
think that is any simpler than the new macro or explicitly testing for
TI/PTI modes.
>>>> +/* { dg-do run } */
>>>> +/* { dg-skip-if "" { powerpc*-*-darwin* } } */
>>>
>>> Nit: This dg-skip-if line looks not necessary as p8vector_hw excludes
>>> *-*-darwin*.
>
> Every single "skip-if darwin" is incorrect if it isn't added by Iain.
> Skipping a test on some target is fine, if it would fail on that target
> (but do put a comment in!), but skipping because you are scared it will
> fail on some target, or you don't care about that target, or just
> cargo-cult, is wrong, and encourages more wrongness.
I added it because all of the other test cases that use
__atomic_compare_exchange*
have a skip-if darwin test and nothing else in the test case looked like it
wasn't supported on darwin, so i thought it must be for the __a_c_e call.
Ok, "all" of the other tests was just pr57744.c, but it's moot anyway, since as
Kewen mentioned, it was redundant due to the p8vector_hw test, so I removed it.
Peter
